[daip] [!4536]: aips - SNSMO in VLMB mode sometimes breaks the relationship between phase(per IF) and delay

[Michael Bietenholz]

Michael Bietenholz updated #4536
--------------------------------

SNSMO in VLMB mode sometimes breaks the relationship between phase(per IF) and delay
------------------------------------------------------------------------------------

           Ticket ID: 4536
                 URL: https://help.nrao.edu/staff/index.php?/Tickets/Ticket/View/4536
           Full Name: Michael Bietenholz
               Email: mbieten at yorku.ca
             Creator: User
          Department: AIPS Data Processing
       Staff (Owner): Eric Greisen
                Type: Issue
              Status: Open
            Priority: Default
                 SLA: NRAO E2E
      Template Group: Default
             Created: 19 February 2014 06:58 PM
             Updated: 27 February 2014 06:03 PM
      Resolution Due: 04 March 2014 07:14 PM (5d 1h 11m)



I think in most uses, which would be to smooth tables from FRING w/ APARM(5)=1, there is actually *one* real SN-table phase value.  The value in e.g., IF1, is a real solution, independent of the multi-band-delay (mbd) solution.  The phase-values in all the other IFs would just be functionally related to the one in IF1 through the mbd.   I usually don't smooth this (single) phase, but it seems conceivable that users may want to.

One option for VLMB would be:  take phase (and rate) from IF1 (or maybe IFmiddle), ignore all others.  Take also mbd.  I  think the (IF-dependent) delays in such cases (ie. FRING:APARM(5)=1) are just equal to the mbd, so you can also ignore the SN-table delay entries and just use the mbd.  Smooth rate,mbd and phase as requested.  After smoothing, fill in the values for the other IFs by just calculating phase from the smoothed IF1 phase and mbd, and copy the smoothed mbd to all the delay slots.  This procedure will produce unpredictable results if the original table has independent IFs, but you could just warn users that they shouldn't use VLMB on such tables.

A second option would be something like de-rotate all phases to IF1 (or IFmiddle) according to the mbd.  Then average phase, rate and delay across IFs.  Then smooth as requested.  Then put IF-dependent phases back.  

I think the first option is likely better, probably easier to implement, and I'm not sure the second option will produce sensible results for SN tables with IF-dependent solutions anyway.   In other words, I'm not sure the whole VLMB mode makes much sense unless you ran a combined-IF FRINGE solution.  I suppose you could implement a quick paranoia-check and find the first non-zero,non-flagged rate & delay in the SN-table.
If the rate is not equal across IFs and delays are not = mbd, then bail.

A potential issue is the SN tables made with APARM(5) = 2,3,4.  Implementing "smoothing over groups of IFs" would be complicated.  

------------------------------------------------------
Staff CP:  https://help.nrao.edu/staff